High-Precision Chip Detection Using YOLO-Based Methods

Na minha lista:
Detalhes bibliográficos
Publicado no:Algorithms vol. 18, no. 7 (2025), p. 448-471
Autor principal: Liu Ruofei
Outros Autores: Zhu Junjiang
Publicado em:
MDPI AG
Assuntos:
Acesso em linha:Citation/Abstract
Full Text + Graphics
Full Text - PDF
Tags: Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!

MARC

LEADER 00000nab a2200000uu 4500
001 3233031667
003 UK-CbPIL
022 |a 1999-4893 
024 7 |a 10.3390/a18070448  |2 doi 
035 |a 3233031667 
045 2 |b d20250101  |b d20251231 
084 |a 231333  |2 nlm 
100 1 |a Liu Ruofei  |u Center for Balance Architecture, Zhejiang University, Hangzhou 310028, China; rf_liu2025@163.com 
245 1 |a High-Precision Chip Detection Using YOLO-Based Methods 
260 |b MDPI AG  |c 2025 
513 |a Journal Article 
520 3 |a Machining chips are directly related to both the machining quality and tool condition. However, detecting chips from images in industrial settings poses challenges in terms of model accuracy and computational speed. We firstly present a novel framework called GM-YOLOv11-DNMS to track the chips, followed by a video-level post-processing algorithm for chip counting in videos. GM-YOLOv11-DNMS has two main improvements: (1) it replaces the CNN layers with a ghost module in YOLOv11n, significantly reducing the computational cost while maintaining the detection performance, and (2) it uses a new dynamic non-maximum suppression (DNMS) method, which dynamically adjusts the thresholds to improve the detection accuracy. The post-processing method uses a trigger signal from rising edges to improve chip counting in video streams. Experimental results show that the ghost module reduces the FLOPs from 6.48 G to 5.72 G compared to YOLOv11n, with a negligible accuracy loss, while the DNMS algorithm improves the debris detection precision across different YOLO versions. The proposed framework achieves precision, recall, and mAP@0.5 values of 97.04%, 96.38%, and 95.56%, respectively, in image-based detection tasks. In video-based experiments, the proposed video-level post-processing algorithm combined with GM-YOLOv11-DNMS achieves crack–debris counting accuracy of 90.14%. This lightweight and efficient approach is particularly effective in detecting small-scale objects within images and accurately analyzing dynamic debris in video sequences, providing a robust solution for automated debris monitoring in machine tool processing applications. 
653 |a Accuracy 
653 |a Video post-production 
653 |a Deep learning 
653 |a Agricultural production 
653 |a Ghosts 
653 |a Machining 
653 |a Neural networks 
653 |a Machine tools 
653 |a Computing costs 
653 |a Algorithms 
653 |a Video data 
653 |a Images 
653 |a Modules 
653 |a Debris 
653 |a Automation 
653 |a Object recognition 
653 |a Morphology 
700 1 |a Zhu Junjiang  |u College of Mechanical and Electrical Engineering, China Jiliang University, Hangzhou 310018, China 
773 0 |t Algorithms  |g vol. 18, no. 7 (2025), p. 448-471 
786 0 |d ProQuest  |t Engineering Database 
856 4 1 |3 Citation/Abstract  |u https://www.proquest.com/docview/3233031667/abstract/embedded/6A8EOT78XXH2IG52?source=fedsrch 
856 4 0 |3 Full Text + Graphics  |u https://www.proquest.com/docview/3233031667/fulltextwithgraphics/embedded/6A8EOT78XXH2IG52?source=fedsrch 
856 4 0 |3 Full Text - PDF  |u https://www.proquest.com/docview/3233031667/fulltextPDF/embedded/6A8EOT78XXH2IG52?source=fedsrch